Angry Citizens and Black Belt Employees: Cascading classifications of and around a predictive algorithm

Abstract

Over past decades, predictive algorithms have been used extensively as profiling tools in the private sector, but today they are also increasingly entering public sector domains. This article builds on an ethnographic study of the development of a predictive algorithm in a debt collecting public sector organization. The algorithm was designed to profile citizens on the basis of their calculated ‘readiness to pay’ their debt and to guide employees’ case handling according to ‘type’ of citizen. The article examines how the classification of citizens produced by the algorithm was mediated by different visualizations and by organizational actors who superimposed new and different classifications (moral and emotional) onto those provided by the algorithm. The article draws on the concepts of nominal and ordinal classification to identify how intended non-hierarchical classification glides into new hierarchical valuations of both citizens and employees. Classifications were ‘cascading’ – a concept the article develops to account for how classification of and around the algorithm multiplied and had organizational ripple effects. Based on empirical insights, the study advocates an agnostic approach to how algorithmic predictions impact work, organizations, and the situation of profiled individuals. It emphasizes a dynamic and and unstable relationship between algorithms and organizational practices.